Category: Frontier AI

This conversation with Ziyue Wang explores some of the most pressing questions at the frontier of AI safety: what it means for powerful AI systems to be safe enough to deploy, how companies model misuse risks such as cyberattacks, CBRN threats, manipulation, loss of control, and malicious actors, and why problems like AI sandbagging remain…